AI, Research & Entrepreneurship

Data Engineer, AI & Data — Riyadh A 28-week co-op with a client delivery team, working on data pipelines and analytics.

• Built end-to-end transformation pipelines in Dataiku DSS, from source data through to published analytical outputs — SQL for set-based work, Python where the logic was procedural
• Validated pipelines against schema contracts, row counts across joins, field-level logic, and boundary conditions
• Worked on performance and maintainability: predicate pushdown, join-order review, removing redundant intermediate datasets, consolidating duplicated logic into shared upstream datasets
• Published validated datasets and dashboards to Power BI, including refresh schedules and connections
• Wrote Low-Level Design documentation that the team adopted as its template, and designed a naming convention and layered taxonomy for pipelines
• Contributed to use-case definition, Point-of-View and PRD documentation, and demos for stakeholders

Selected work

• Built a proof of concept as a working application, turning data the system already held into ranked recommendations; it supported the team in securing a new project
• Proposed a use case based on an observed gap; the client adopted it for development
• Contributed to the design of an anomaly-detection approach using per-entity time-series forecasting, chosen because the signal was relative rather than absolute and no labelled data existed
• Proposed an incremental-processing layer to skip unchanged records before expensive model inference

The first intelligent monitoring system built to measure and analyse average student attention in Saudi classrooms. FOCAL turns raw classroom video into structured attention metrics — and turns passive cameras into an instructor-support tool that alerts in real time when attention drops.

• Five-module pipeline: frame sampling (1 frame / 5 seconds — over 90% less computation), YOLOv8 person detection, ByteTrack multi-person tracking, SSD face detection with Gaussian blurring, and cloud behaviour classification via Roboflow
• The FOCAL Formula — a weighted average that turns five classroom behaviours (raising a hand, focusing, looking to a friend, using a phone, sleeping) and their confidence scores into one interpretable attention score, with a minimum-frame threshold to filter noise
• Privacy-first by design: every face is blurred before analysis, body posture and gestures preserved; no individual scoring, no surveillance, no emotion recognition — class-level analysis only
• Outputs overall, highest and lowest attention to live and historical dashboards, plus LLM-generated flashcards and exportable reports, on a Supabase backend
• Delivered across four sprints in a 14-week semester · Showcased at the PSU Senior Project CCIS Expo
• Supervisors: Dr. Zinab Masood & Dr. Abrar Wafa

AI-powered smart glasses delivering offline, real-time diagnostic support to first responders in crisis-affected regions.

• First Place + Exceptional Innovation Award — King Salman Relief & Alfaisal University Hackathon 2025
• Featured at the UNDP-led Human Capital Initiative Dialogue
• Team lead — with the support of Dr. Noor Alsaadoun

An AI-powered chatbot that makes Saudi cultural heritage accessible through interactive, culturally grounded conversation. Existing solutions for visitors to the Kingdom either lacked localized content or offered no dynamic, immersive experience — Suhail was built to close that gap.

• Two novel datasets created for this work, derived from trusted and verified sources — 93.3% accuracy
• Validated through expert interviews and survey responses
• Evolved from a course project into a publication and a functional product
• Presented at the Women in Data Science (WiDS) Conference
• Serving Saudi Vision 2030 and UN SDGs 4 (Quality Education) & 11 (Sustainable Cities) — Supervisor: Dr. Heyfa Ammar

A study of how commuting and occupational experience shape mental well-being among 813 Saudi adults, measured on the WHO-5 Well-Being Index — the first work to combine emotional and logistical commute data for predictive modeling in a non-Western context.

• Dense Neural Network regression (64 → dropout → 32 → 1), Adam optimizer, MSE loss, 200 epochs with early stopping and a 20% validation split
• R² = 0.25 — a substantial effect size by current psychological research standards (r ≈ 0.5)
• SHAP analysis identified the dominant predictors: commute mood first, then job satisfaction and vehicle comfort — all outranking commute duration. How we feel on the way matters more than how far we travel.
• Presented at the 8th Women in Data Science (WiDS) International Conference @ Prince Sultan University
• Co-authored with Dr. Abrar Wafa, College of Computer & Information Sciences

Project manager of a high-performing AI model detecting visual pollutants (potholes and more).

• 77.71 Mean Average Precision on unseen images
• Active learning pipeline for near-automatic labeling; transfer learning, TTA, cascade ensembling

Led research on AI's impact on national identities and cultural communication, focused on Saudi Arabia.

• Finding: 73.2% of participants believe intelligent technologies can protect cultural heritage
• Outputs: "The Kingdom's Journey in AI" booklet + an interactive cultural AI application

The project that started it all. 41% of adolescents with anxiety have no access to mental health services — and the classical route is expensive, clinical, and weighed down by stigma. So I embedded the assessment itself inside a game: a dialogue of in-game choices that gathers diagnostic data while the player simply plays.

• Adapted the Taylor Manifest Anxiety Scale — dry true/false clinical items — into choose-your-path scenarios inside Minecraft: Story Mode, designed to be added to any game adolescents already play daily
• Community survey of 543 participants aged 14–19 exposed the real gap: 79.7% wanted a diagnosis, but only 5.3% had ever been diagnosed by a specialist — the reluctance is about how diagnosis is delivered, not whether it's wanted
• Blind diagnostic trial with 35 adolescents: the game-based diagnosis proved statistically similar to the standard scale (p = 0.0552) — accurate and credible, at lower cost and with greater reach
• 3rd Place worldwide, Behavioral & Social Sciences — Regeneron ISEF 2021 + American Psychological Association Award
• Featured in Smithsonian Magazine, Innovation for Good: Game Changers

• The Big Digital Pitch — First Place (intercollegiate final of 10, PSU, 2023): a solution for OCD and the social stigma around it
• Monsha'at Academy — "From Idea to Innovation" bootcamp (2021)
• SDA Hemmah Game Development BootCamp — youngest in the Kingdom's first batch (2021)
• Saudi Leadership Society delegation to Shenzhen, China (Dec 2025): Huawei, Tencent, DJI, Meituan, Unilumin — studying how innovation ecosystems let ideas survive and scale

• Computer Vision Internship — Robotics & Internet-of-Things (RIOTU) Lab (2025), with Dr. Abrar Wafa and Prof. Wadii Boulila
• Talmatha Research Program — Mawhiba's competitive program for gifted researchers (2020)